Purpose: To ensure that the system or deployment is beneficial to human well-being and avoids harms.
| Organisational / Technical | Measure |
|---|---|
| A. Integrate features that detect threats to well-being | |
| TECH | Automated classifiers allow detecting topics or behaviours that indicate risks to well-being |
| TECH | Implement crisis detection AI that flags up possible critical situations in the health or well-being of users |
| B. Prevent uses that may jeopardize user well-being | |
| TECH | Integrate keyword filtering to catch out of scope requests |
| TECH | Constraint training data to limit model knowledge to appropriate domains |
| TECH | Monitor outputs to prevent harmful responses even if filters missed request |
| C. Reporting of well-being impacts | |
| BOTH | Set up reporting system and customer team interface to let users report on well-being impacts |
Source: AIOLIA deliverable 3.1